What it means
An NPU (Neural Processing Unit) is a processor specialized for neural-network operations, chiefly the multiply-accumulate math of running trained models. Like other ASICs it trades general-purpose flexibility for efficiency on a narrow workload, delivering more inference per watt than a CPU and, for its target tasks, often better efficiency than a GPU. NPUs appear both as blocks inside phones, PCs, and other devices for on-device AI and as standalone data-center accelerators. In the AI supply chain they represent the custom-silicon alternative to merchant GPUs, letting device makers and cloud operators tailor hardware to inference. NPUs are a lever for pushing inference cost down and closer to the user, but a constraint in that they still depend on the same foundries, packaging, and memory as the rest of the accelerator stack.
Why it matters to investors
NPUs spread the AI trade beyond the data center into phones and PCs, and give buyers a custom-silicon route around merchant GPUs for inference workloads. The design and foundry ecosystem behind custom accelerators captures that shift.
Companies on this part of the chain
Named to show where the term sits in the AI supply chain — research, not advice, and never a recommendation to buy or sell.
Related terms
See NPU in the live AI chain.
THE ENTITY maps every constraint onto one live model — which part is tight now, who owns it, and who gets squeezed when it moves. Plain-English reads you can check.
THE ENTITY is an educational read on the AI supply chain — research, not investment advice. It explains how the chain works and who sits where, never price targets or buy/sell calls.